stBoost)Geiser C. Challco geiser@alumni.usp.br
cond <- "stBoost"
to_remove <- c('S11')
sub.groups <- c("age","ed.level","intervention","age:intervention",
"ed.level:intervention","age:ed.level:intervention")dat <- read_excel("../data/data-without-outliers.xlsx", sheet = "fss-cond-descriptive")
dat <- dat[!dat$study %in% to_remove, ]
leg <- read_excel("../data/data-without-outliers.xlsx", sheet = "legend")## New names:
## • `` -> `...10`leg <- leg[!leg$study %in% to_remove, ]
idx.e <- which(dat$condition==cond)
idx.c <- which(dat$condition=="control")
data <- data.frame(
study = dat$study[idx.c],
n.e = dat$N[idx.e], mean.e = dat$M.emms[idx.e], sd.e = dat$SD.emms[idx.e],
n.c = dat$N[idx.c], mean.c = dat$M.emms[idx.c], sd.c = dat$SD.emms[idx.c]
)
for (cgroups in strsplit(sub.groups,":")) {
data[[paste0(cgroups, collapse = ":")]] <- sapply(data$study, FUN = function(x) {
paste0(sapply(cgroups, FUN = function(namecol) leg[[namecol]][which(x == leg$study)]), collapse = ":")
})
}
data[["lbl"]] <- sapply(data$study, FUN = function(x) leg$Note[which(x == leg$study)])m.cont <- metacont(
n.e = n.e, mean.e = mean.e, sd.e = sd.e, n.c = n.c, mean.c = mean.c, sd.c = sd.c,
studlab = lbl, data = data, sm = "SMD", method.smd = "Hedges",
fixed = F, random = T, method.tau = "REML", hakn = T, title = paste("Performance in",cond)
)
summary(m.cont)## Review: Performance in stBoost
##
## SMD 95%-CI %W(random)
## S1 0.6017 [ 0.0897; 1.1136] 10.1
## S2 0.0200 [-0.4562; 0.4962] 11.6
## S3 -0.1774 [-0.7134; 0.3587] 9.2
## S4 0.5120 [-0.0442; 1.0682] 8.5
## S5 0.0845 [-0.3315; 0.5005] 15.3
## S6 0.3476 [-0.1021; 0.7973] 13.1
## S7 0.1612 [-0.2355; 0.5580] 16.8
## S10: Only use prompt msgs -0.0106 [-0.4241; 0.4028] 15.4
##
## Number of studies combined: k = 8
## Number of observations: o = 597
##
## SMD 95%-CI t p-value
## Random effects model 0.1741 [-0.0312; 0.3793] 2.01 0.0849
##
## Quantifying heterogeneity:
## tau^2 < 0.0001 [0.0000; 0.2358]; tau = 0.0006 [0.0000; 0.4856]
## I^2 = 8.8% [0.0%; 70.4%]; H = 1.05 [1.00; 1.84]
##
## Test of heterogeneity:
## Q d.f. p-value
## 7.67 7 0.3625
##
## Details on meta-analytical method:
## - Inverse variance method
## - Restricted maximum-likelihood estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Hedges' g (bias corrected standardised mean difference; using exact formulae)forest(m.cont, digits=2, digits.sd = 2, test.overall = T, label.e = cond)
m.sg4sub <- update.meta(m.cont, subgroup = age, random = T, fixed = F)
summary(m.sg4sub)## Review: Performance in stBoost
##
## SMD 95%-CI %W(random) age
## S1 0.6017 [ 0.0897; 1.1136] 10.1 adolescent
## S2 0.0200 [-0.4562; 0.4962] 11.6 adolescent
## S3 -0.1774 [-0.7134; 0.3587] 9.2 adolescent
## S4 0.5120 [-0.0442; 1.0682] 8.5 adult
## S5 0.0845 [-0.3315; 0.5005] 15.3 adult
## S6 0.3476 [-0.1021; 0.7973] 13.1 adult
## S7 0.1612 [-0.2355; 0.5580] 16.8 adult
## S10: Only use prompt msgs -0.0106 [-0.4241; 0.4028] 15.4 adolescence
##
## Number of studies combined: k = 8
## Number of observations: o = 597
##
## SMD 95%-CI t p-value
## Random effects model 0.1741 [-0.0312; 0.3793] 2.01 0.0849
##
## Quantifying heterogeneity:
## tau^2 < 0.0001 [0.0000; 0.2358]; tau = 0.0006 [0.0000; 0.4856]
## I^2 = 8.8% [0.0%; 70.4%]; H = 1.05 [1.00; 1.84]
##
## Test of heterogeneity:
## Q d.f. p-value
## 7.67 7 0.3625
##
## Results for subgroups (random effects model):
## k SMD 95%-CI tau^2 tau Q I^2
## age = adolescent 3 0.1498 [-0.8478; 1.1473] 0.0917 0.3028 4.71 57.5%
## age = adult 4 0.2406 [-0.0406; 0.5218] 0 0 1.83 0.0%
## age = adolescence 1 -0.0106 [-0.4241; 0.4028] -- -- 0.00 --
##
## Test for subgroup differences (random effects model):
## Q d.f. p-value
## Between groups 1.25 2 0.5343
##
## Details on meta-analytical method:
## - Inverse variance method
## - Restricted maximum-likelihood estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Hedges' g (bias corrected standardised mean difference; using exact formulae)forest(m.sg4sub, digits=2, digits.sd = 2, test.overall = T, label.e = cond)
m.sg4sub <- update.meta(m.cont, subgroup = ed.level, random = T, fixed = F)
summary(m.sg4sub)## Review: Performance in stBoost
##
## SMD 95%-CI %W(random) ed.level
## S1 0.6017 [ 0.0897; 1.1136] 10.1 upper-secundary
## S2 0.0200 [-0.4562; 0.4962] 11.6 upper-secundary
## S3 -0.1774 [-0.7134; 0.3587] 9.2 upper-secundary
## S4 0.5120 [-0.0442; 1.0682] 8.5 higher-education
## S5 0.0845 [-0.3315; 0.5005] 15.3 higher-education
## S6 0.3476 [-0.1021; 0.7973] 13.1 higher-education
## S7 0.1612 [-0.2355; 0.5580] 16.8 unknown
## S10: Only use prompt msgs -0.0106 [-0.4241; 0.4028] 15.4 upper-secundary
##
## Number of studies combined: k = 8
## Number of observations: o = 597
##
## SMD 95%-CI t p-value
## Random effects model 0.1741 [-0.0312; 0.3793] 2.01 0.0849
##
## Quantifying heterogeneity:
## tau^2 < 0.0001 [0.0000; 0.2358]; tau = 0.0006 [0.0000; 0.4856]
## I^2 = 8.8% [0.0%; 70.4%]; H = 1.05 [1.00; 1.84]
##
## Test of heterogeneity:
## Q d.f. p-value
## 7.67 7 0.3625
##
## Results for subgroups (random effects model):
## k SMD 95%-CI tau^2 tau Q I^2
## ed.level = upper-secundary 4 0.1025 [-0.4182; 0.6233] 0.0397 0.1993 5.10 41.2%
## ed.level = higher-education 3 0.2767 [-0.2494; 0.8028] 0 0 1.60 0.0%
## ed.level = unknown 1 0.1612 [-0.2355; 0.5580] -- -- 0.00 --
##
## Test for subgroup differences (random effects model):
## Q d.f. p-value
## Between groups 0.78 2 0.6761
##
## Details on meta-analytical method:
## - Inverse variance method
## - Restricted maximum-likelihood estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Hedges' g (bias corrected standardised mean difference; using exact formulae)forest(m.sg4sub, digits=2, digits.sd = 2, test.overall = T, label.e = cond)
m.sg4sub <- update.meta(m.cont, subgroup = intervention, random = T, fixed = F)
summary(m.sg4sub)## Review: Performance in stBoost
##
## SMD 95%-CI %W(random)
## S1 0.6017 [ 0.0897; 1.1136] 10.1
## S2 0.0200 [-0.4562; 0.4962] 11.6
## S3 -0.1774 [-0.7134; 0.3587] 9.2
## S4 0.5120 [-0.0442; 1.0682] 8.5
## S5 0.0845 [-0.3315; 0.5005] 15.3
## S6 0.3476 [-0.1021; 0.7973] 13.1
## S7 0.1612 [-0.2355; 0.5580] 16.8
## S10: Only use prompt msgs -0.0106 [-0.4241; 0.4028] 15.4
## intervention
## S1 Gender-stereotype color, ranking, badges, and avatar
## S2 Gender-stereotype color, ranking, badges, and avatar
## S3 Gender-stereotype color, ranking, badges, and avatar
## S4 Gender-stereotype color, ranking, badges, and avatar
## S5 Gender-stereotype color, ranking, badges, and avatar
## S6 Gender-stereotype color, ranking, badges, and avatar
## S7 Gender-stereotype color, ranking, badges, and avatar
## S10: Only use prompt msgs Gender-stereotyped motivational message prompts
##
## Number of studies combined: k = 8
## Number of observations: o = 597
##
## SMD 95%-CI t p-value
## Random effects model 0.1741 [-0.0312; 0.3793] 2.01 0.0849
##
## Quantifying heterogeneity:
## tau^2 < 0.0001 [0.0000; 0.2358]; tau = 0.0006 [0.0000; 0.4856]
## I^2 = 8.8% [0.0%; 70.4%]; H = 1.05 [1.00; 1.84]
##
## Test of heterogeneity:
## Q d.f. p-value
## 7.67 7 0.3625
##
## Results for subgroups (random effects model):
## k SMD 95%-CI tau^2 tau Q I^2
## intervention = Gender-stereotype color, rankin ... 7 0.2078 [-0.0265; 0.4421] <0.0001 0.0013 6.76 11.3%
## intervention = Gender-stereotyped motivational ... 1 -0.0106 [-0.4241; 0.4028] -- -- 0.00 --
##
## Test for subgroup differences (random effects model):
## Q d.f. p-value
## Between groups 0.89 1 0.3457
##
## Details on meta-analytical method:
## - Inverse variance method
## - Restricted maximum-likelihood estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Hedges' g (bias corrected standardised mean difference; using exact formulae)forest(m.sg4sub, digits=2, digits.sd = 2, test.overall = T, label.e = cond)
m.sg4sub <- update.meta(m.cont, subgroup = `age:intervention`, random = T, fixed = F)
summary(m.sg4sub)## Review: Performance in stBoost
##
## SMD 95%-CI %W(random)
## S1 0.6017 [ 0.0897; 1.1136] 10.1
## S2 0.0200 [-0.4562; 0.4962] 11.6
## S3 -0.1774 [-0.7134; 0.3587] 9.2
## S4 0.5120 [-0.0442; 1.0682] 8.5
## S5 0.0845 [-0.3315; 0.5005] 15.3
## S6 0.3476 [-0.1021; 0.7973] 13.1
## S7 0.1612 [-0.2355; 0.5580] 16.8
## S10: Only use prompt msgs -0.0106 [-0.4241; 0.4028] 15.4
## age:intervention
## S1 adolescent:Gender-stereotype color, ranking, badges, and avatar
## S2 adolescent:Gender-stereotype color, ranking, badges, and avatar
## S3 adolescent:Gender-stereotype color, ranking, badges, and avatar
## S4 adult:Gender-stereotype color, ranking, badges, and avatar
## S5 adult:Gender-stereotype color, ranking, badges, and avatar
## S6 adult:Gender-stereotype color, ranking, badges, and avatar
## S7 adult:Gender-stereotype color, ranking, badges, and avatar
## S10: Only use prompt msgs adolescence:Gender-stereotyped motivational message prompts
##
## Number of studies combined: k = 8
## Number of observations: o = 597
##
## SMD 95%-CI t p-value
## Random effects model 0.1741 [-0.0312; 0.3793] 2.01 0.0849
##
## Quantifying heterogeneity:
## tau^2 < 0.0001 [0.0000; 0.2358]; tau = 0.0006 [0.0000; 0.4856]
## I^2 = 8.8% [0.0%; 70.4%]; H = 1.05 [1.00; 1.84]
##
## Test of heterogeneity:
## Q d.f. p-value
## 7.67 7 0.3625
##
## Results for subgroups (random effects model):
## k SMD 95%-CI tau^2 tau Q
## age:intervention = adolescent:Gender-stereotype co ... 3 0.1498 [-0.8478; 1.1473] 0.0917 0.3028 4.71
## age:intervention = adult:Gender-stereotype color, ... 4 0.2406 [-0.0406; 0.5218] 0 0 1.83
## age:intervention = adolescence:Gender-stereotyped ... 1 -0.0106 [-0.4241; 0.4028] -- -- 0.00
## I^2
## age:intervention = adolescent:Gender-stereotype co ... 57.5%
## age:intervention = adult:Gender-stereotype color, ... 0.0%
## age:intervention = adolescence:Gender-stereotyped ... --
##
## Test for subgroup differences (random effects model):
## Q d.f. p-value
## Between groups 1.25 2 0.5343
##
## Details on meta-analytical method:
## - Inverse variance method
## - Restricted maximum-likelihood estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Hedges' g (bias corrected standardised mean difference; using exact formulae)forest(m.sg4sub, digits=2, digits.sd = 2, test.overall = T, label.e = cond)
m.sg4sub <- update.meta(m.cont, subgroup = `ed.level:intervention`, random = T, fixed = F)
summary(m.sg4sub)## Review: Performance in stBoost
##
## SMD 95%-CI %W(random)
## S1 0.6017 [ 0.0897; 1.1136] 10.1
## S2 0.0200 [-0.4562; 0.4962] 11.6
## S3 -0.1774 [-0.7134; 0.3587] 9.2
## S4 0.5120 [-0.0442; 1.0682] 8.5
## S5 0.0845 [-0.3315; 0.5005] 15.3
## S6 0.3476 [-0.1021; 0.7973] 13.1
## S7 0.1612 [-0.2355; 0.5580] 16.8
## S10: Only use prompt msgs -0.0106 [-0.4241; 0.4028] 15.4
## ed.level:intervention
## S1 upper-secundary:Gender-stereotype color, ranking, badges, and avatar
## S2 upper-secundary:Gender-stereotype color, ranking, badges, and avatar
## S3 upper-secundary:Gender-stereotype color, ranking, badges, and avatar
## S4 higher-education:Gender-stereotype color, ranking, badges, and avatar
## S5 higher-education:Gender-stereotype color, ranking, badges, and avatar
## S6 higher-education:Gender-stereotype color, ranking, badges, and avatar
## S7 unknown:Gender-stereotype color, ranking, badges, and avatar
## S10: Only use prompt msgs upper-secundary:Gender-stereotyped motivational message prompts
##
## Number of studies combined: k = 8
## Number of observations: o = 597
##
## SMD 95%-CI t p-value
## Random effects model 0.1741 [-0.0312; 0.3793] 2.01 0.0849
##
## Quantifying heterogeneity:
## tau^2 < 0.0001 [0.0000; 0.2358]; tau = 0.0006 [0.0000; 0.4856]
## I^2 = 8.8% [0.0%; 70.4%]; H = 1.05 [1.00; 1.84]
##
## Test of heterogeneity:
## Q d.f. p-value
## 7.67 7 0.3625
##
## Results for subgroups (random effects model):
## k SMD 95%-CI tau^2 tau
## ed.level:intervention = upper-secundary:Gender-stereoty ... 3 0.1498 [-0.8478; 1.1473] 0.0917 0.3028
## ed.level:intervention = higher-education:Gender-stereot ... 3 0.2767 [-0.2494; 0.8028] 0 0
## ed.level:intervention = unknown:Gender-stereotype color ... 1 0.1612 [-0.2355; 0.5580] -- --
## ed.level:intervention = upper-secundary:Gender-stereoty ... 1 -0.0106 [-0.4241; 0.4028] -- --
## Q I^2
## ed.level:intervention = upper-secundary:Gender-stereoty ... 4.71 57.5%
## ed.level:intervention = higher-education:Gender-stereot ... 1.60 0.0%
## ed.level:intervention = unknown:Gender-stereotype color ... 0.00 --
## ed.level:intervention = upper-secundary:Gender-stereoty ... 0.00 --
##
## Test for subgroup differences (random effects model):
## Q d.f. p-value
## Between groups 1.46 3 0.6921
##
## Details on meta-analytical method:
## - Inverse variance method
## - Restricted maximum-likelihood estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Hedges' g (bias corrected standardised mean difference; using exact formulae)forest(m.sg4sub, digits=2, digits.sd = 2, test.overall = T, label.e = cond)
m.sg4sub <- update.meta(m.cont, subgroup = `age:ed.level:intervention`, random = T, fixed = F)
summary(m.sg4sub)## Review: Performance in stBoost
##
## SMD 95%-CI %W(random)
## S1 0.6017 [ 0.0897; 1.1136] 10.1
## S2 0.0200 [-0.4562; 0.4962] 11.6
## S3 -0.1774 [-0.7134; 0.3587] 9.2
## S4 0.5120 [-0.0442; 1.0682] 8.5
## S5 0.0845 [-0.3315; 0.5005] 15.3
## S6 0.3476 [-0.1021; 0.7973] 13.1
## S7 0.1612 [-0.2355; 0.5580] 16.8
## S10: Only use prompt msgs -0.0106 [-0.4241; 0.4028] 15.4
## age:ed.level:intervention
## S1 adolescent:upper-secundary:Gender-stereotype color, ranking, badges, and avatar
## S2 adolescent:upper-secundary:Gender-stereotype color, ranking, badges, and avatar
## S3 adolescent:upper-secundary:Gender-stereotype color, ranking, badges, and avatar
## S4 adult:higher-education:Gender-stereotype color, ranking, badges, and avatar
## S5 adult:higher-education:Gender-stereotype color, ranking, badges, and avatar
## S6 adult:higher-education:Gender-stereotype color, ranking, badges, and avatar
## S7 adult:unknown:Gender-stereotype color, ranking, badges, and avatar
## S10: Only use prompt msgs adolescence:upper-secundary:Gender-stereotyped motivational message prompts
##
## Number of studies combined: k = 8
## Number of observations: o = 597
##
## SMD 95%-CI t p-value
## Random effects model 0.1741 [-0.0312; 0.3793] 2.01 0.0849
##
## Quantifying heterogeneity:
## tau^2 < 0.0001 [0.0000; 0.2358]; tau = 0.0006 [0.0000; 0.4856]
## I^2 = 8.8% [0.0%; 70.4%]; H = 1.05 [1.00; 1.84]
##
## Test of heterogeneity:
## Q d.f. p-value
## 7.67 7 0.3625
##
## Results for subgroups (random effects model):
## k SMD 95%-CI tau^2 tau
## age:ed.level:intervention = adolescent:upper-secundary:Gend ... 3 0.1498 [-0.8478; 1.1473] 0.0917 0.3028
## age:ed.level:intervention = adult:higher-education:Gender-s ... 3 0.2767 [-0.2494; 0.8028] 0 0
## age:ed.level:intervention = adult:unknown:Gender-stereotype ... 1 0.1612 [-0.2355; 0.5580] -- --
## age:ed.level:intervention = adolescence:upper-secundary:Gen ... 1 -0.0106 [-0.4241; 0.4028] -- --
## Q I^2
## age:ed.level:intervention = adolescent:upper-secundary:Gend ... 4.71 57.5%
## age:ed.level:intervention = adult:higher-education:Gender-s ... 1.60 0.0%
## age:ed.level:intervention = adult:unknown:Gender-stereotype ... 0.00 --
## age:ed.level:intervention = adolescence:upper-secundary:Gen ... 0.00 --
##
## Test for subgroup differences (random effects model):
## Q d.f. p-value
## Between groups 1.46 3 0.6921
##
## Details on meta-analytical method:
## - Inverse variance method
## - Restricted maximum-likelihood estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Hedges' g (bias corrected standardised mean difference; using exact formulae)forest(m.sg4sub, digits=2, digits.sd = 2, test.overall = T, label.e = cond)
m.cont <- update.meta(m.cont, studlab = data$study)
summary(eggers.test(x = m.cont))## Eggers' test of the intercept
## =============================
##
## intercept 95% CI t p
## 2.745 -3.54 - 9.03 0.856 0.42
##
## Eggers' test does not indicate the presence of funnel plot asymmetry.funnel(m.cont, xlab = "Hedges' g", studlab = T, legend=T, addtau2 = T)